Public opinions of light rail service in Los Angeles, an analysis using Twitter data

New ways to understand opinion data in Twitter in term of sentiment analysis, topic modeling, and the interaction between posters and retweeters are introduced, which could contribute to improved situational understanding, crisis communications, enhancing and delivering transport policy goals.

Abstract

Understanding commuters' perceptions, attitudes, and behavior is an important component of transportation planning and management. Collecting such information using traditional survey or interview methods is costly and burdensome, but mining attitudinal data from social networking media could potentially provide insights into the temporal alignment of public opinion with transportation system dynamics. We demonstrate this potential by examining facets of public posts on Twitter about light rail transit services in Los Angeles in terms of sentiment analysis, topic modeling, and the interaction between posters and retweeters. Results provide new insights into how transit users present themselves and their opinions, engage with government agencies, react to events/policies, and share information with others on social media. We also demonstrate an interactive online interface that transit service providers could use to display and monitor real-time feedback and sentiment along different lines in the area's light rail system. With the explosive growth of social web services and mobile devices, individuals and organizations are increasingly using information from social networking media for decision-making. Twitter is a form of blogging that allows users to send brief text updates or micromedia such as photographs or audio clips. Through the service, users provide information on geo-location (GPS coordinates from smartphones), timing of where and when they provided information, and their opinions/reviews. Several social sectors (e.g. politics, entertainment and business) have been gathering and analyzing information from Twitter (Mittal & Goel, 2012; Tumasjan, Sprenger, Sandner, & Welpe, 2010). Using information from this data source could reduce the need to conduct surveys, opinion polls, and focus groups because there is such abundance of information is already publicly available. However, finding opinions on Twitter and distilling the information into meaningful patterns is a time consuming task. The average human reader will have difficulty identifying relevant sites and extracting and summarizing the opinions from large numbers of daily Twitter posts. We developed an analysis system based on text mining techniques to address this need. Though using Twitter data for opinion mining is popular in many fields, its use in transportation planning and management sector remains still limited. In 2013, Iteris (Mai & Hranac, 2013) compared incident records from the California Highway Patrol with Twitter messages related to roadway events over the same time period. In another study, Twitter data were obtained from the riders of the rapid transit system of the Chicago Transit Authority which allows members of Twitter to monitor sentiments of transit users (Collins, Hasan, & Ukkusuri, 2012). However, these studies are still far from sufficient public opinion mining. This paper extends these previous efforts by introducing ways to understand opinion data in Twitter in term of sentiment analysis, topic modeling, and the interaction between posters and retweeters, which could contribute to improved situational understanding, crisis communications, enhancing and delivering transport policy goals. We conduct detailed analysis for each line in 7-line rail transit system in Los Angeles.